Carbonatite Fluid Identification Method based on fuzzy C-means clustering
Technical field
The invention belongs to petroleum exploration field, relate to the chaos quantum particle swarm optimization algorithm is combined with the fuzzy C-means clustering algorithm, and be incorporated in the middle of the carbonatite fluid identification, realize that for taking full advantage of multiple prestack elastic information the carbonatite fluid identification provides a kind of new research thinking.
Background technology
Utilizing seismic data to carry out fluid identification is the most important research work of oil-gas exploration, because prestack AVO/AVA inverting keeps the shear wave information on stratum, the reservoir fluid variation characteristic is had susceptibility, so the AVO technology has become one of important earth physical means of fluid detection.
Some scholars have done a large amount of deep research work to utilizing the prestack elastic information to carry out fluid identification both at home and abroad, have proposed to utilize the multiple sensitivity fluid identification factor to come convection cell directly to detect (Goodway, 1997; Ning Zhonghua, 2006; Li Jingye, 2008; He Zhenhua, 2009), obtained effect preferably in some areas.But be difficult in actual applications select the suitable fluid identification factor according to actual conditions, particularly at Tarim Basin carbonate complex rock reservoir, it is unsatisfactory to utilize the single fluid identification factor to carry out fluid predicted application effect.Maturation and effective method is that the rock physics analysis is combined with the prestack elastic inversion relatively at present, the preferred responsive fluid identification factor crosses to carry out the prediction of fluid.At first, carry out the fluid substitution studies by meticulous rock physics analysis, the various prestack elastic parameters in the forward simulation reservoir during saturated different fluid character (oil, gas, water); Estimate various elastic parameters then to the sensitivity of reservoir lithology and fluid, analyze preferred two kinds of responsive fluid identification combinations of factors by crossing; At last, as a result on the basis, the corresponding responsive fluid factor of extracting at the well lie crosses, and irises wipe out the scope of full gas, full oil and full water reservoir, thereby draws fluid distribution integrated forecasting result (Jiang Wei, 2010 in the study area in the prestack elastic inversion; Lang Xiaoling, 2010; Wang Jun, 2011).
This method is compared with the poststack Fluid Identification Method very big progress, but the practical application effect in the carbonatite fluid identification of Tarim Basin but is not very desirable.Analyze it and mainly contain following some reason: 1. rock physics research is the basic guarantee of carrying out prestack inversion, and Tarim Basin carbonate reservoir secondary pores structure (corrosion hole, hole, seam) complexity, this can bring sizable difficulty to the rock physics analysis of carbonate reservoir and the prediction of p-and s-wave velocity, inevitably bring error in computation process, its order of accuarcy can play significant effects to the prestack inversion result; 2. rock physics is just being drilled the log data frequency higher (being generally 1kHZ-20kHz) of analyzing institute's foundation, and often there is big frequency dispersion in seismic data and causes frequency lower (being generally 10HZ-80Hz) in deep layer, can there be certain difference in log data and inversion result on numerical range, therefore be from the prestack inversion result, to extract in the actual data of utilizing when carrying out fluid identification that cross, it is different that the scope that is full of different fluid character that causes drawing a circle to approve in the responsive fluid factor crosses figure is just being drilled the scope of drawing a circle to approve in the analysis with rock physics, in order to obtain fluid identification effect preferably, must demarcate according to the fluid information that fixed well discloses, constantly revise the fluid distribution range among the figure that crosses, to reach best matching effect,, unavoidably bringing artifical influence factor in the operating process like this; 3. can calculate multiple prestack elastic parameter by prestack AVO/AVA inverting, wherein comprised abundant fluid information, more sensitive two kinds of prestack parameters cross but the final preferred two kinds of convection cells identification of conventional prestack Fluid Identification Method compares, thereby convection cell distributes and predicts, and fail to take full advantage of abundant prestack information, cause the precision of fluid identification not high.
Summary of the invention
Responsive and easily be absorbed in the shortcoming of local convergence to initial value at fuzzy C-means clustering (FCM) algorithm, the present invention organically combines FCM algorithm and chaos quantum population (CQPSO) algorithm, a kind of fuzzy C-means clustering based on the chaos quantum population (CQPSO-FCM) method has been proposed, this method utilizes the chaos particle cluster algorithm to come initialization degree of membership matrix, can effectively solve the fuzzy C-means clustering algorithm to the initialization sensitive issue, and have the ability of very strong search globally optimal solution, effectively improved the fuzzy classification ability.And improved fuzzy C-means clustering algorithm is incorporated in the middle of the carbonatite fluid identification, extract fluid properties that known drilling well discloses and the relation between the prestack elastic parameter, can calculate the probability distribution that sample to be distinguished belongs to each fluid properties.Verified, but the uncertainty of this method convection cell identification is estimated, can significantly improve the precision of carbonatite fluid identification, thereby can reduce exploration risk, have good actual application and be worth, realize that for taking full advantage of the prestack elastic information identification of carbonate complex rock flow body provides a kind of new technical method.
The present invention realizes that the specific embodiments of above-mentioned purpose is as follows:
Step 1: seismic data is carried out prestack protect width of cloth skew, extract common reflection point CRP road collection;
Step 2: carry out prestack AVO inverting, based on inversion result: p-wave impedance, shear wave impedance, density data body calculate the multiple fluid factor;
Step 3: the fluid properties actual conditions that disclose according to drilling well, choose study area drilling well as training sample, and extract the corresponding fluid factor of each well reservoir section and cross in twos, according to susceptibility and independency principle, preferred convection cell property identification sensitivity, can reflect the fluid properties essential characteristic and each other independently as responsive fluid identification combinations of factors;
Step 4: the degree of membership matrix that utilizes chaotic maps initialization fuzzy C-means clustering, preferred responsive fluid identification combinations of factors is analyzed in higher dimensional space, calculate each cluster centre and sample apart from each distances of clustering centers, the fluid properties of training of judgement sample, when satisfying termination condition, change step 6 over to, otherwise change step 5 over to;
Step 5: utilize quantum particle swarm that the degree of membership matrix of fuzzy C-means clustering is upgraded iteration, and whether evaluation algorithm is absorbed in precocious convergence, if then carry out chaotic maps, up to finding global optimum, make the prediction rate of coincideing of training sample fluid properties satisfy termination condition, change step 6 over to;
Step 6: fuzzy clustering number and each fuzzy clustering center of exporting each fluid properties;
Step 7: calculate sample to be identified apart from distance and the degree of membership at each fuzzy clustering center, each fluid properties will be belonged to respectively, degree of membership as oil gas, water, mud and matrix adds up, and calculates the probability that sample to be identified belongs to each fluid properties, identifies as foundation convection cell character with this.
Fuzzy C-means clustering based on the chaos quantum population provided by the present invention (CQPSO-FCM) algorithm basic principle is as follows: responsive and easily be absorbed in the shortcoming of local convergence to initial value at fuzzy C-means clustering algorithm (FCM), utilize chaotic maps initialization degree of membership matrix, can effectively solve the fuzzy C-means clustering algorithm to the initialization sensitive issue; And utilize chaos particle cluster algorithm (CQPSO) to replace the iterative process of fuzzy C-means clustering algorithm (FCM), constantly quanta particle is upgraded by the calculating target function error, the judgment mechanism of the precocious convergence of operation simultaneously, if target function value does not satisfy termination condition, and colony's fitness variance restrains judgment threshold less than precocity, think that then population is absorbed in local extremum, introduce the chaos mutation operation, preferred optimum mapping point replaces relatively poor particle in the former population from chaos sequence, significantly strengthen the ability of the search globally optimal solution of algorithm, effectively improved the fuzzy classification ability.
Characteristic of the present invention is: the multiple prestack elastic parameter that is finally inversed by by prestack AVO has comprised abundant fluid information, but conventional prestack Fluid Identification Method finally only preferred two kinds of convection cells identification more sensitive fluid identification factor that compares cross, and fail to take full advantage of abundant prestack information, cause the precision of fluid identification not high.For some in the inseparable data of two dimensional surface, and be easy to separately at higher dimensional space, from this thought, improved Fuzzy C homogeneous clustering algorithm is incorporated in the middle of the carbonatite fluid identification, in higher dimensional space, set up the fluid properties of the announcement of drilling well and the relation between the multiple responsive fluid identification factor, namely can efficient solution determine because rock physics analysis result and the unmatched problem of seismic inversion result that the frequency dispersion of seismic data causes, can fully merge multiple prestack elastic parameter again, improve the accuracy of identification of carbonatite fluid.And this method not only can obtain the classification of fluid properties to be judged, and can also calculate the degree of membership (being the fluid class probability) that belongs to each fluid properties, but the uncertainty of convection cell identification estimate, thereby can effectively reduce exploration risk.
Description of drawings
Fig. 1 is based on the carbonatite Fluid Identification Method schematic flow sheet of fuzzy C-means clustering
Fig. 2 extracts the well lie multiple fluid factor synoptic diagram that crosses
The preferred responsive fluid identification combinations of factors higher dimensional space perspective view of Fig. 3
Fig. 4 is based on each fluid properties cluster centre higher dimensional space perspective view of fuzzy C-means clustering analytical calculation
Fig. 5 X5 (well) and X5C (oil gas well) improve fuzzy C-means clustering fluid identification result schematic diagram
Fig. 6 X9 (shale filling well) improves fuzzy C-means clustering fluid identification result schematic diagram
Certain block of area is based on the fluid distribution characteristics synoptic diagram of fuzzy C-means clustering prediction in the tower of Fig. 7 Tarim Basin
Embodiment
Below in conjunction with the description of drawings specific embodiment of the invention.
Fig. 1 is the carbonatite Fluid Identification Method schematic flow sheet based on fuzzy C-means clustering of the present invention:
Step 1: seismic data is carried out prestack protect width of cloth skew, extract common reflection point CRP road collection;
Step 2: carry out prestack AVO inverting, based on inversion result: p-wave impedance, shear wave impedance, density data body calculate the multiple fluid factor;
Step 3: the fluid properties actual conditions that disclose according to drilling well, choose study area drilling well as training sample, and extract the corresponding fluid factor of each well reservoir section and cross in twos, according to susceptibility and independency principle, preferred convection cell property identification sensitivity, can reflect the fluid properties essential characteristic and each other independently as responsive fluid identification combinations of factors;
Step 4: the degree of membership matrix that utilizes chaotic maps initialization fuzzy C-means clustering, preferred responsive fluid identification combinations of factors is analyzed in higher dimensional space, calculate each cluster centre and sample apart from each distances of clustering centers, the fluid properties of training of judgement sample, when satisfying termination condition, change step 6 over to, otherwise change step 5 over to;
Step 5: utilize quantum particle swarm that the degree of membership matrix of fuzzy C-means clustering is upgraded iteration, and whether evaluation algorithm is absorbed in precocious convergence, if then carry out chaotic maps, up to finding global optimum, make the prediction rate of coincideing of training sample fluid properties satisfy termination condition, change step 6 over to;
Step 6: fuzzy clustering number and each fuzzy clustering center of exporting each fluid properties;
Step 7: calculate sample to be identified apart from distance and the degree of membership at each fuzzy clustering center, the degree of membership that belongs to each fluid properties (oil gas, water, mud and matrix) is respectively added up, calculate the probability that sample to be identified belongs to each fluid properties, identify as foundation convection cell character with this.
Data such as comprehensive logging explanation, formation testing are determined the fluid properties of each well in the carbonate reservoir section, and hypothesis is consistent with pithead position apart from well head 50m scope inner fluid character, choose 12 mouthfuls of wells that study area inner fluid character is relatively determined, extract the fluid factor of 1246 data points in the other 50m scope of well as known training sample, wherein oil and gas reservoir is 195,95 of moisture reservoirs, 65 of shale filling reservoirs, 891 of non-reservoirs.Utilize the fluid properties of algorithm predicts training sample provided by the invention, and compare with known fluid character, calculate coincidence rate, check the prediction effect of this algorithm fluid properties.Predict the outcome and show that fuzzy C-means clustering is all quite remarkable to the classifying quality of each fluid properties, the coincidence rate of oil and gas reservoir and non-reservoir has all reached more than 95%, total coincidence rate reaches 97.99%, illustrate that this method can take full advantage of abundant prestack information, well distinguish the difference between each fluid properties, verified the validity of this method.
Fig. 2 is the fluid properties actual conditions that disclose according to drilling well, extracts the corresponding fluid factor of each well reservoir section and crosses in twos.Choosing in the tower of Tarim Basin certain block of area is study area, and the fluid properties actual conditions that drilling well discloses according to study area are divided into oil and gas reservoir, moisture reservoir, shale filling reservoir and non-reservoir four classes with carbonate reservoir.Result's (p-wave impedance, shear wave impedance, density data body) based on prestack inversion can calculate the multiple fluid factor, according to the recognition effect to the study area fluid properties, comprise at the fluid factor of utilizing among study area the present invention: p-wave impedance Ip, p-and s-wave velocity are than the product λ ρ of Vp/Vs, Lame's constant and density and μ ρ and two kinds of combination parameter λ ρ * VpVs and Ip*VpVs.And these six kinds of fluid factors are crossed in twos, therefrom preferably to the fluid combinations of factors of carbonatite fluid identification sensitivity.
Fig. 3 is the projection of preferred responsive fluid identification combinations of factors higher dimensional space.Fig. 2 multithread body factor result that crosses is analyzed owing to be not separate between each fluid factor, the correlativity between must the analysing fluid factor, optimize can reflect the fluid properties essential characteristic, the independent fluid factor each other.Wherein, Vp/Vs and λ ρ, Vp/Vs and λ ρ * Vp/Vs, all has good correlativity between Ip*Vp/Vs and the Ip, reject the bigger fluid factor of coefficient that is relative to each other simultaneously, keep Vp/Vs, Ip and μ ρ at last as responsive fluid identification combinations of factors, and carry out projection at three dimensions, and it is easier separately than any two kinds of fluid factors fluid properties that crosses as seen to cross at three dimensions, and the recognition capability of the sensitive factor combination convection cell that namely the multidimensional MD is little is stronger.
Fig. 4 is based on each fluid properties cluster centre higher dimensional space projection of fuzzy C-means clustering analytical calculation.Utilization is analyzed preferred Vp/Vs, Ip and the responsive fluid identification combinations of factors of μ ρ based on the fuzzy C-means clustering algorithm of chaos quantum particle group optimizing, because the distribution range difference of different fluid character in three dimensions, therefore different fluid character has different fuzzy clustering numbers, calculate the fuzzy clustering center of each fluid properties at last, well express the distribution characteristics of the fluid factor of different fluid character in three dimensions, reflected the essential characteristic of each fluid properties in the prestack elastic information.
Fig. 5 is that X5 (well) improves fuzzy C-means clustering fluid identification result with X5C (oil gas well).X5 well (well) and X5C well (oil gas well) are though all be shown as " beading strong reflection " on stacked seismic data, but fluid properties has very big difference, the well completing test of straight well X5 well is mainly to produce water, it is flammable to light a fire, test result is the gassiness water layer, and sidetracked hole X5C test obtains the high yield commercial hydrocarbon flow, can calculate the degree of membership (being the fluid class probability) that belongs to different fluid character by improving the fuzzy C-means clustering method, the moisture probability of straight well X5 well is the highest as can be seen, and the oily probability of sidetracked hole X5C well is the highest, therefore the X5 well is classified as the product well, the X5C well is classified as the oil gas well, and is consistent with practical condition.
Fig. 6 is that X9 (shale filling well) improves fuzzy C-means clustering fluid identification result.The X9 well is the typical shale filling of study area well, and test contains oil bloom, and the shale filling is serious, and test result is for doing layer, contains with this well of prediction that the mud probability is the highest to coincide.
Fig. 7 is the fluid distribution characteristics that certain block of area is predicted based on fuzzy C-means clustering in the tower of Tarim Basin, the characteristic of fluid that discloses with drilling well all well coincide, prove that it is effective that this method is carried out the carbonatite fluid identification, ability with fine differentiation fluid, can fully merge multiple prestack elastic information the carbonatite fluid is comprehensively identified, can improve the fluid identification precision of carbonatite.And this method not only can obtain the distribution characteristics of fluid properties, and can also draw the degree of membership (being the fluid class probability) that belongs to each fluid properties, but the uncertainty of convection cell identification is estimated, thereby can effectively reduce exploration risk, has good actual application and is worth.
Above embodiment only is used for explanation the present invention, but not is used for limiting the present invention.